Rebuilt Jira and Confluence sign-up by cutting friction, one experiment at a time.

As the designer on Atlassian’s sign-up team, my job was simple to say and hard to do: get more new teams into Jira and Confluence, and using them in their first week.

The funnel was leaking everywhere. 75% of people who started a Jira sign-up never landed in the product. The flow had grown to 10 screens of choices, questions and confirmations. Three steps stood out:

• Email verification: sending people to their inbox to click a link lost 52% of Jira and 73% of Confluence sign-ups.
• Name your site: 29% of Jira and 51% of Confluence users dropped here. People didn’t know what a “site” was and were wary of naming something permanent for their company.
• Choosing a bundle: an extra decision before people had even seen the product.

If I can get up and going with a trial within a minute or two of pressing that first button, that’s awesome. If it takes me 10 minutes just to verify my email and put a password in, I’m probably not going to spend as much time.

Research participant

One early experiment cut 20% of the steps at once. Sign-ups jumped, but a bug created duplicate accounts and muddied what had actually caused the lift. We couldn’t ship it, and the lesson stuck: we needed to change one thing at a time to know what worked.

Friction or motivation? The team had two theories. One said people dropped off because sign-up felt far from the product, so we should show product visuals, progress bars and overlays to keep them motivated. The other said people dropped off because we kept asking them to do things. I pushed for the second: remove steps first, add polish only where it helped someone finish.

That path was harder. Email verification and site naming belonged to other teams, and changing them meant negotiating access to their code and earning their trust.

Verification: finding the version we could ship. Skipping email verification entirely was tested with users and they didn’t mind, but security put it on hold. With the Identity team, we landed on a one-time passcode typed into the page instead of a link in a separate inbox. It kept the security check and removed the trip away from sign-up.

Site name: the big swing, with downstream risk. Removing the site name step looked like the biggest win. But a site is how teams add more Atlassian products later, and new sites drove 36% of Jira purchases. I mapped every place a user would meet an auto-generated name like “team-38092” later: welcome emails, admin settings, adding a second product. Given how much was still unknown, I suggested we split the bet. We ran a lower-risk, customer-facing experiment that recommended a site name instead of asking for one, while engineering tested the hardest parts of full removal behind the scenes. The recommendation won. The removal test surfaced too many downstream issues, so we didn’t ship it.

Looking outside. I audited sign-up and onboarding at 20+ companies, including Notion, Linear, Figma and Monday. It showed most strong flows asked for less, used plain words like “company name,” and skipped decorative product shots. PMs and the Head of Product used it to shape the next quarter’s roadmap.

Each experiment changed one thing, so we knew exactly what earned each lift.

Experiment	What changed	Result
One-time passcode	A 6-digit code on the page instead of a link in a separate email	+11.8% Jira and +16.7% Confluence first-week activation
Recommended site name	A suggested name instead of a blank, permanent-feeling field	+5.7% Jira and +11.4% Confluence sign-up completion
Remove bundles	Cut the “pick a bundle” screen before sign-up	Significant lift; about 144 more active business sites a month for Jira
Simplified sign-up	Removed decorative and repeated content	+10.6% Confluence sign-ups; neutral for Jira, giving a clean base to test on
Jobs to be done	People picked tasks instead of a template	Mixed: more day-one use for software and marketing teams, a drop for Jira Work Management
Product visuals behind sign-up	Sign-up floated over an image of the product	Flat; rolled back

All results above were statistically significant except where noted as flat or mixed.

Keeping tests from colliding. With many experiments running across sign-up, onboarding and Identity, tests started overlapping. In one quarter, five experiments had to restart because they changed the same screens or showed the wrong flow. I mapped every live experiment by launch date and the steps it changed, so the team could see collisions before launch and sequence around them.

Craft where it earned its place. I kept polish focused on moments that helped people finish: progress feedback and success states, animated in After Effects so we didn’t wait on an outside animator. I also scoped back animation in one experiment to prove the idea first, saving engineering about a month.

Over six months, the share of people who started sign-up and were active in their first week rose 47.5% for Jira and 89% for Confluence. Sign-up owned the first half of that journey; we worked closely with the Onboarding team on the rest.

• 10 sign-up experiments designed in six months, against a goal of 7
• Email verification: the share of people getting past this step rose from about 40% to 67.5%, helped by the passcode and faster email delivery
• Team goals: 106% of the Confluence target for new active business sites, 96% of Jira’s
• Across the Jira sign-up redesign (a team-wide effort I contributed to): 11 experiments shipped, adding about 6,100 active business sites in the fiscal year

What didn’t work. The motivation bets fell flat. Product visuals and overlays behind sign-up didn’t move sign-ups or activation and were rolled back. That confirmed the direction: removing friction beat decorating it.

In the 6 months we worked together, she helped spearhead the vision for Atlassian Sign Up, launched 9+ experiments, and improved D1-D6/sign up start rates significantly for JSW and Confluence.

Alisha Manandhar, Product Manager

Next steps

• About a third of people still dropped at verification, so there was more to remove there
• Bring the site name and project naming into a single, plain “company name” step
• Use the jobs-to-be-done answers to personalize onboarding, with the Onboarding team
• Focus on teammates joining an existing site, which drove over half of sign-ups